Shallow-trench isolation (STI) is currently the most popular isolation scheme for advanced CMOS (e.g. 0.25 μm and beyond) due to its superior capability for minimum isolation pitch, better planar surface, and enhanced latch-up immunity. However, as CMOS technology continuously advances to 0.13 μm and beyond, the STI isolation spacing is small (e.g. ˜0.1 μm, where μ=10−6) enough for significant field penetration into the transistor channel (or body) area from adjacent poly gates as well as drain/source junctions. As a result of such electrical field penetration, there is a depletion region formed along the channel side-wall of the STI causing transistor narrow width effect and transistor Vt fluctuations. Please refer to the following publications:
A. Bryant, W. Hänsch, and T. Mii, “Characteristics of CMOS Device Isolation for the ULSI Age”, International Electron Device Meetings, p. 671, 1994.
C. Wang, P. Zhang, “Three-Dimensional DIBL for Shallow-Trench Isolation MOSFET's”, IEEE Trans. on Electron Device, V. 46, No. 1, p. 139, 1999.
J. H. Sim, J. K. Lee, and K. Kim, “The Impact of Isolation Pitch Scaling on VTH Fluctuation in DRAM Cell Transistors due to Neighboring Drain/Source Electric Field Penetration”, Symposium of VLSI technology, p. 32-33, 1998.
In one way to eliminate such field penetration effect, a shielded STI, is proposed by J. H. Sim, J. K. Lee, and K. Kim, in “High-performance cell transistor design using metallic shield embedded shallow trench isolation 9MSE-STI) for Gbit generation DRAM's”, IEEE Transaction on Electron Devices, Vol. 46, No. 6, p. 1212-1217, 1999 by filling a layer of conducting material (e.g. doped poly) after liner oxidation of the isolation trench. The grounded conducting material in the trench can provide good shielding and therefore eliminate the transistor narrow-width effect and Vt fluctuations.
As illustrated in FIG. 1 (not to scale), the n-channel body region 15a in p-substrate 12 will have a depletion layer 17 formed along the side-walls of STI 16 as a result of fringe field (as indicated by arrows A) from both poly-gate 19 and adjacent n-doped regions 18 or from adjacent poly gate (not shown), when biased high (e.g. +Vcc). The fringe field from the poly gate edge and n-doped regions (biased high) can expel holes and form depletion layers 17 along the side-wall of STI.
Similarly, a corresponding p-channel body region 15b in n-well 14 will also have a depletion layer 17 formed by the fringe field (as indicated by arrows A) from the edge of poly-gate 19 and adjacent p-doped regions 20 as shown in FIG. 2 (not to scale). Similarly, the fringe field from the poly gate edge and p-doped regions (biased low) can expel electrons and form depletion layers along the side-wall of STI. A few parameters of typical 0.13 μm CMOS technology are listed here for reference: STI depth ˜0.4 μm, STI minimum width 0.1-0.15 μm, n-doped or p-doped region depth ˜0.8-0.12 μm.
STI field penetration effect is considered undesirable in advanced CMOS (especially in DRAM, or SPAM arrays), a metallic-shielded STI was proposed by J. H. Sim, J. K. Lee, and K. Kim in “High-performance cell transistor design using metallic shield embedded shallow trench isolation (MSE-STI) for Gbit generation DRAM's”, IEEE Transaction on Electron Devices, Vol. 46, No. 6, p. 1212-1217, 1999, by filling a layer of conducting material (e.g. doped poly) after liner oxidation of the trench. The grounded conducting material in the trench can provide a good shielding and therefore can eliminate Vt fluctuations by the field penetration effect.
A well-known circuit configuration of a voltage reference is described by Gray and Meyer, “Analog VLSI Circuit Analysis”, chapter 12, Wiley, 1984, and is illustrated in FIG. 10 using 2 n-MOSFFTs with different threshold voltages of Vt1 and Vt2, referred to as a “VT-difference ” voltage reference circuit. The two n-MOSFETs of the same size (i.e. same W/L) are biased by the same magnitude (typically 0.1 μA to 100 μA) current source I, and both n-MOSFETs are used as “pull-up” transistors. The threshold voltage Vt of the two transistors is made different by either channel implant or by a different doping type of the poly gate. The gate of the first n-MOSFET with Vt1 is grounded (as a convenient reference voltage). An operational amplifier (op-amp) OA is connected to the source sides A and B (for detecting the difference of Vt) and the op-amp output Vo is connected to the gate of the second n-MOSFET with Vt2 (for maintaining the second n-MOSFET turn-on). The output Vo from the op-amp is simply the Vt difference of the two n-MOSFET transistors, i.e. Vt1-Vt2. The accuracy of the circuits depends on the size matching of MOS transistors and the offset of the op-amp. The basic circuit configuration in FIG. 10 can be modified by various additional circuits for trimming or calibration purposes, and is widely used in CMOS VLSI. The temperature coefficient of this circuit can be very good due to the cancellation of temperature dependence of n-MOSFET with different Vt. Correspondingly, a reference circuit using two p-MOSFETs can be similarly implemented.
Related U.S. patents pertinent to the invention are:
U.S. Pat. No. 6,078,094 (Poplevine et al.) shows a variable width vertical resist and STI process.
U.S. Pat. No. 6,051,474 (Beasom) teaches a method to bias the isolation trench fill.
U.S. Pat. No. 5,899,724 (Dobuzinsky et al.) describes a TIN vertical resistor. However, this reference differs from the invention.
U.S. Pat. No. 5,234,861 (Roisen et al.) discloses a method to form an isolation structure and to optionally bias it.
U.S. Pat. No. 4,933,739 (Harari) describes a vertical trench resistor.
The undesirable field penetration effect through STI in advanced CMOS can be utilized for a class of new vertical (variable) resistor and FET structures by biasing adjacent junctions for depletion layer or accumulation layer formation in the resistor region along the side-wall of an STI. This new family of devices (vertical resistor and FET) can be formed by CMOS compatible technology.
Also disclosed is a new voltage reference provided by utilizing such vertical FET with n+ and p+ control junctions. The difference in Fermi-levels of n+ and p+ doping (i.e. band-gap) in the control junction is used in the Vt-difference circuit configuration and the output is simply one silicon band-gap. Compared with conventional MOS transistor Vt-difference voltage reference, the proposed voltage reference has smaller layout (due to the vertical nature of FET) with less contact and connections, and smaller temperature coefficient.